Hey guys, quick update on the external "math brain" I've been building for AI quoting agents.
I hit a massive edge case: what happens when the customer haggles or changes their mind? If you let standard LLMs recalculate the price mid-chat, they instantly hallucinate.
To fix this, I built an inline chatbot directly into the quote flow that acts as a live salesman.
Here is the HVAC example from the Loom:
- Customer: "I need to reduce the price."
- Negotiation: The AI suggests options. Customer replies: "Let's do flexible 7-14 day scheduling to drop the $350 rush fee."
- The Ping: Instead of guessing, the bot pings the Quotix math engine.
- The Update: The engine does the strict math and live-updates the quote on the screen to $11,500.
No hallucinated discounts, no broken math.
Here are two quick Looms showing it in action:
- The Inline Chatbot (HVAC negotiation): [Inline Chatbot Loom Link]
- The Backend Architecture: [Architecture Loom Link]
Would love your brutal feedback. How are you guys handling it when a customer tries to change a quote mid-conversation?